Stata statistical program

Diet costs were calculated as per the methods of the Healthy Diets ASAP protocol [17 (link)], where the collected food and drink prices were applied to the 75 representative food and drinks. The mean and standard error of intake for household members in SEGs were calculated using survey weights with the Stata statistical program [26 ]. The mean and standard error of intakes for a household in an SEG was calculated assuming the independence of intakes of household members. Fixed-effects meta-regression models were used to perform a test for the linear trend of intakes across SEGs. p-values ≤ 0.05 were considered significant.
The cost of the recommended diet was calculated using the recommended diet pricing tool of the Healthy Diets ASAP protocol [17 (link)]. As healthy, equitable, and more sustainable dietary recommendations are generally similar for all Australians, the recommended diet cost was the same for each SEG.
The affordability of both habitual and recommended diets was calculated for each SEG using the calculated quintile income ranges. Diet costs were deemed unaffordable if they were 30% or more of the household disposable income [17 (link)]. If 25–29%, the household was considered to be in food stress [27 (link)].

Multivariable logistic regression was used to identify correlates of social violence. A binary dependent variable representing any social violence (0 = No, 1 = Yes) was constructed based on students’ statements regarding their experience of three different forms of social violence; this dependent variable took the value 1 if a participant had experienced any social violence in the past month. Independent variables included province, district (rural/urban), age, parents’ highest education, self-assessed level of masculinity/femininity compared to other members of the same sex (less/equally/more masculine/feminine), self-described gender/sexual identity, grade point average (GPA), access to the Internet on a mobile phone/computer, having a steady/casual partner in the past year, alcohol consumption, and substance use (e.g. cannabis, amphetamine pills, crystal methamphetamine, ecstasy/MDMA, sleeping pills/benzodiazepines or injection drug use).
Two multivariable logistic regression models were created: Model 1 included students’ self-assessed masculinity/femininity and other factors and model 2 included gender/sexual identity and other factors. Both models were adjusted for age, GPA, parents’ highest education, province and district. STATA statistical program (Version 14.0, StataCorp, College Station, Texas) was used to perform all statistical analyses.